Network Theoretic Analysis of Maximum a Posteriori Detectors for Optimal Input Detection

TL;DR

This paper analyzes the performance of MAP detectors for detecting changes in network inputs using noisy sensor data, providing explicit performance characterizations and sensor placement insights based on network structure.

Contribution

It explicitly characterizes MAP detector performance in network settings and offers structural insights into optimal sensor placement based on detection theory.

Findings

Performance depends on network edge weights and sensor locations.

Conditions identified where sensor placement improves detection.

Validation through numerical examples confirms theoretical results.

Abstract

This paper considers maximum-a-posteriori (MAP) and linear discriminant based MAP detectors to detect changes in the mean and covariance of a stochastic input, driving specific network nodes, using noisy measurements from sensors non-collocated with the input nodes. We explicitly characterize both detectors' performance in terms of the network edge weights and input and sensor nodes' location. In the asymptotic measurement regime, when the input and measurement noise are jointly Gaussian, we show that the detectors' performance can be studied using the input to output gain of the system's transfer function matrix. Using this result, we obtain conditions for which the detection performance associated with the sensors on a given network cut is better (or worse) than that of the sensors associated with the subnetwork induced by the cut and not containing the input nodes. Our results also provide structural insights into the sensor placement from a detection-theoretic viewpoint. We validate our theoretical findings via multiple numerical examples.